<html>
<head>
<meta content="text/html; charset=windows-1252"
http-equiv="Content-Type">
</head>
<body bgcolor="#FFFFFF" text="#000000">
<div class="moz-cite-prefix"><a class="moz-txt-link-freetext" href="http://mailinglists.scilab.org/Scilab-6-0-0-has-been-released-tp4035500p4035570.html">http://mailinglists.scilab.org/Scilab-6-0-0-has-been-released-tp4035500p4035570.html</a><br>
<br>
Le 22/02/2017 à 21:50, Amanda Osvaldo a écrit :<br>
</div>
<blockquote type="cite">
<meta http-equiv="content-type" content="text/html;
charset=windows-1252">
<br>
<b>* ParFor</b><br>
<p style="margin-bottom: 0cm; line-height: 100%">Forward
parallelism,
it would also be very convenient to be able to parallelize
operations
within the macro without having to put the code to be paralyzed
in
another macro.</p>
<p style="margin-bottom: 0cm; line-height: 100%">For example:</p>
<p style="margin-bottom: 0cm; line-height: 100%">data = [1:10];</p>
<p style="margin-bottom: 0cm; line-height: 100%">r = [];</p>
<p style="margin-bottom: 0cm; line-height: 100%"><b>parfor</b> i =
1:10</p>
<p style="margin-bottom: 0cm; line-height: 100%"> r(i) = data(i) *
5;
// Each operation will be performed in parallel</p>
<p style="margin-bottom: 0cm; line-height: 100%">endfor</p>
<p style="margin-bottom: 0cm; line-height: 100%">// You will only
reach this line after all the parallel operations have finished.</p>
<p style="margin-bottom: 0cm; line-height: 100%">disp(r);</p>
<title></title>
<meta name="generator" content="LibreOffice 5.1.5.2 (Windows)">
<style type="text/css">
@page { margin: 2cm }
h1 { margin-bottom: 0.21cm }
h1.western { font-family: "Liberation Sans", sans-serif; font-size: 18pt }
h1.cjk { font-family: "Droid Sans Fallback"; font-size: 18pt }
h1.ctl { font-family: "Droid Sans Devanagari"; font-size: 18pt }
p { margin-bottom: 0.25cm; line-height: 120% }
a:link { so-language: zxx }
</style></blockquote>
<br>
Indeed, this would be a very handy and straightforward way to run a
parallelized loop.<br>
<br>
For the time being, although its help page is still distributed,<br>
<a class="moz-txt-link-freetext" href="https://help.scilab.org/docs/6.0.0/en_US/parallel_run.html">https://help.scilab.org/docs/6.0.0/en_US/parallel_run.html</a><br>
<b><i>parallel_run</i></b> has been removed from Scilab 6<br>
<tt>--> parallel_run</tt><tt><br>
</tt><tt>Undefined variable: parallel_run</tt><tt><br>
</tt><br>
It was pointed to one year ago: <a class="moz-txt-link-freetext" href="http://bugzilla.scilab.org/14486">http://bugzilla.scilab.org/14486</a><br>
<br>
And some MPI features announced in the restricted 6.0.0 release
notes are not actually distributed:<br>
<i>* MPI_Create_comm create a new communicator from MPI_COMM_WORLD
using MPI world ranks.</i><i><br>
</i><tt><br>
--> MPI_Create_comm</tt><tt><br>
</tt><tt>Impossible to load mpi.dll library</tt><tt><br>
</tt><br>
at least on the MSWindows 64 bits plateform, for which the Scilab
installer proposes by default to instal the MPI module.<br>
<br>
So, parallelization is somewhat regressing. On windows, not really,
since parallel_run never actually addressed more than one processor.<br>
<br>
Samuel<br>
<br>
</body>
</html>